Comment 4 for bug 1381340

Revision history for this message
JuanJo Ciarlante (jjo) wrote :

As we found later, this was most likely due to an MTU misconfiguration: hosts' interfaces set to mtu=9000 while switches not fully setup for it - after getting them properly setup, this deployment went ok.

Nevertheless, having meaningful messages at juju log would have helped a lot, warnings/errors like e.g. 'read timeout from tcp connection'.